--MD tag log failure

[clalogiudice]

Hello everybody,
I am trying to map my Pacbio reads with minimap2.
Tryiing to add the flag --MD causes a failure in the minimap alignment log...

[M::mm_idx_gen::859.8700.51] collected minimizers
[M::mm_idx_gen::882.2310.83] sorted minimizers
[M::main::882.2310.83] loaded/built the index for 93 target sequence(s)
[M::mm_mapopt_update::886.5150.83] mid_occ = 663
[M::mm_idx_stat] kmer size: 15; skip: 5; is_hpc: 0; #seq: 93
[M::mm_idx_stat::889.631*0.83] distinct minimizers: 167026627 (35.39% are singletons); average occurrences: 5.939; average spacing: 3.163
minimap2: forminimap2: format.c:190: write_MD_core: Assertion `op >= 0 && op <= 2' failed.

[lh3]

--MD doesn't work with splice alignment because I don't know how to encode introns in MD.

[tambua]

Thank you, just another question....
Does the "samtools calmd " command consider introns as indels?

[lh3]

I don't know. Although I implemented the initial calmd, I have forgotten the details and the samtools team may have made significant improvements to it. You have to ask the samtools team unfortunately.

[tambua]

Thank you for your quick answer, I'll ask them!

[lh3]

I have seen your question at samtools/samtools#809. Do you prefer MD to include introns or not to include introns? What's the use case in your mind? Thank you.

[SaberHQ]

@lh3 With --MD tag not working with splice alignment, we should expect --cs tag not to work too? As you are constructing --cs out of --MD/cigar and the query/ref sequences. Am I right?

I see no error when I use --cs with splice alignment, but the output sam files do not contain cs tags. It would be great if you can produce a warning to let users know about it.

[kehey]

I get the same error with the --MD flag, which is now clear of course. But on the comment above by @SaberHQ: I do get the cs tags in the SAM output when running a spliced alignment with --cs

[lh3]

@SaberHQ --cs has been working for a long time. If it doesn't, please send me an example.

3a119d6 enables --MD to work with spliced alignment. I am closing this issue. If the new code has bugs, please create a new issue.